Methodology

Gyanis is built on a calibration engine that adapts across Flow, Vibe, and Aura. This page explains the methods behind the system so that researchers, educators, and regulators can understand how we make recommendations to students.

Data sources

• Official syllabus documents for each exam board (CBSE, NCERT, AQA, Edexcel, OCR, KPM, KICD, UAE MOE).
• Past paper archives licensed or publicly available through the respective exam board.
• Teacher-authored content commissioned for topic weightings, common mistakes, and marking scheme interpretation.
• Aggregate, non-identifying user signals (answered questions, time on task, topic mastery) to calibrate adaptive difficulty. Identifying fields are removed before any signal enters the calibration layer.

The calibration engine

1. Knowledge tracing: ELO-style difficulty bands track what each student actually knows at the level of individual sub-topics.
2. Next-best action:A multi-armed bandit policy picks whether to drill a weak area, revise a recent topic, or push into new material, subject to the student's daily time budget.
3. Confidence smoothing: Nightly smoothing separates a rough single session from a real gap in understanding.
4. Human override:Teachers using Gyanis can always override the system's recommendations for their students.

Model accuracy

We publish internal benchmark reports twice a year showing how well Flow's difficulty predictions match actual student performance on held-out past paper items. Aggregate numbers are shared on the Gyanis blog; detailed reports are available to schools under a data sharing agreement.

Aura (wellbeing) — not algorithmic

Aura sessions are conducted by qualified psychologists; the platform does not use AI to diagnose, advise on medication, or replace professional clinical judgment. We do use automated mood prompts and session booking, clearly labelled as such.

Transparency commitments

• We do not use student data to train third-party models without explicit opt-in.
• We do not serve interest-based ads to students; we do not share student data with advertisers.
• Parents can request a copy of their child's data or deletion at any time via privacy@gyanis.ai.